The Impact of Software Design on Energy Performance

Software consumes energy as it executes, and the amount of energy expended on its execution globally is considerable. A constant growth in device ownership combined with a shift to cloud computing and a growing awareness of climate change has led to a new focus on software performance, but now from an energy perspective. The issue of software energy efficiency can be approached from several angles --- this thesis explores the relationship between software design and energy efficiency. Contemporary object-oriented programming practices prioritise modularity, extensibility and reusability in an effort to develop well-organised and maintainable code. However the introduction of the structures and indirection inherent to these goals results in redundancy, with the potential for significant and unnecessary energy consumption. This work investigates refactoring, design patterns, software metrics, and code smells to develop an understanding of the impact of high-level software design on energy efficiency. Studies undertaken in this work achieve energy consumption reductions of between 5% and 66% when large, open-source applications are transformed to reduce energy consumption, and also presents a novel approach to estimating energy performance by weighting static metrics with dynamic profiling data. These results underscore the negative impact software design practices can have on energy efficiency and raise questions regarding maintainability versus energy efficiency. The findings in this thesis also develop our understanding of how software design can impact energy efficiency, guides future research in this space, and enables developers to make informed decisions when developing more energy-aware software.